Pulse width modulation vs efficiency

[PeterScowcroft]

Hi,

I have been basing my design off the datasheets for the XR-E/ Seoul P4 and may have made an oversight.

When the Seoul is running at 1A the efficiency is ~60lm/w and ~100lm/w at 350mA.

I have been told the graph might compensate for heat output.

So,

If a seoul is pulsed on a 1/3 cycle of 1A will the efficiency still be 60lm/W or will it climb to be nearer teh 100lm/w?


Is there a technical help line for Seoul anyone knows?

 

[uk_caver]

Though there will be some heat component to the efficiency loss with increased [average] current, white LED efficiencies at very low currents are often extraordinarily high, but start dropping long before currents are reached at which the device is likely to warm up significantly, and also drop fastest at low powers, so there's likely something more than heat at work.

As a first approximation, it's probably best to assume that for a given 'on' current, the efficiency of the LED is the same as its efficiency would be at that current applied constantly.

 

[RustyKnee]

interesting question.

leds have less colour shift whem pwmed compared to a constant current switcher. I guess this indicates that the led behaves similarly to at constant fulll current.

Stu

 

[uk_caver]

I know there's been a deal of work making phosphors more efficient at high temperatures, but I don't know whether they are also simply more efficient at low levels of illumination.
With lumens depending on the light colour, I guess any decrease in the fraction of blue light converted to yellow could be a significant lumens hit even if the blue die in the LED had a perfectly linear relationship between current and light output.

 

[PeterScowcroft]

Hmm good point.

I have e-mailed Seoul direct but while waiting for them to get back does anyone else agree/disagree with the above?

 

[PeterScowcroft]

Hmm good point.

I have e-mailed Seoul direct but while waiting for them to get back does anyone else agree/disagree with the above?

 

[CM]

Efficiency is predominantly influenced by heat. When you PWM a LED, you need to compute the average power at the particular duty cycle. From there you can go to the data sheet to determine what the efficiency is by looking at output vs. bias current.

 

[wasBlinded]

Efficiency is predominantly influenced by heat. When you PWM a LED, you need to compute the average power at the particular duty cycle. From there you can go to the data sheet to determine what the efficiency is by looking at output vs. bias current.

That is not correct. There is a loss of conversion efficiency at increasing currents that is independent of die temperature. An LED driven at 1 amp at a 25% duty cycle will have a much lower efficiency than the same LED driven at 250 mA and a 100% duty cycle, so your stated method will overestimate lumen production. That isn't to deny the role of heat, but heat isn't the whole story by any means.

For more details, find Newbie's posts and graphs (extensive) on the subject.

 

[PeterScowcroft]

So am I assuming by this CM that with an exceptional cooling system the efficiency could be dramatically increased?

Or does that chart take into account the maximum speed teh die can vent heat through the thermal path?

I am thinking that if you can run them bright and efficient at a high current if cooled...wait for it...the worlds best freezer light!!!!

 

[kakster]

Courtesy of Jarhead_Newbie:

http://www.molalla.net/~leeper/pwm_cc.htm

 

[uk_caver]

Looking at the Seoul and Cree datasheets, the Seoul sheet has the light-vs-current graph at a constant ambient temperature (without making it exactly obvious what kind of heatsink performance was involved), but the Cree datasheet seems to suggest their curve is at a constant junction temperature, and both curves look broadly similar in terms of light fall-off with current. Luxeon datasheets also seem to use a constant Tj, but then they don't show the low end of the graphs.

 

[PeterScowcroft]

cool thx,
I will post Seouls responce if/when they get back to me.

 

[fineday]

I think it will be more than 60LM but less than 100LM, because the temperature will be lower than working at 1A constant current but when it's working, the current is 1A.

Just my own opinion.

 

[RustyKnee]

Courtesy of Jarhead_Newbie:

http://www.molalla.net/~leeper/pwm_cc.htm

Very intersting stuff

Stu

 

[Coop]

http://www.molalla.net/~leeper/seoulmy.htm

Check out the last chart. Again thanks to Jarhead_newbie....

 

[RustyKnee]

http://www.molalla.net/~leeper/seoulmy.htm

Check out the last chart. Again thanks to Jarhead_newbie....

What does that last chart show in context to this thread?

it just seem to show that the a p4 seoul is slightly more effieicient than a p4 xre.


Stu

 

[lightrod]

Observation: battery life on a PWM light is linearly related (inversely) to output. Determine the runtime for a given cell type at a given output. Runtime at 1/4 that output is then seen to be 4x.

I've seen this on my FF3 (lux 3) as well as my Draco (XRE). Example for the Draco is here -

At full output (level 10) life is very short and I do not think the curve is very stable - perhaps at that point the heat is in fact a factor. However once a definable/stable curve is attained (e.g. at level 7) the conclusion above holds very true - the area under the curves is almost identical.

This is indirect, but IMO compelling, evidence that you get no meaningful efficiency improvement for a PWM light as you reduce output.

In general then, for PWM dimming you expect cell life to increase 2x for a 1/2 cut in output; for current dimming you might get about a 3x improvement in cell life for a 1/2 cut in output. That's the downside of the PWM dimming method - you always get the relatively poor efficiency that's related to the full current.

 

[CM]

That is not correct. There is a loss of conversion efficiency at increasing currents that is independent of die temperature. An LED driven at 1 amp at a 25% duty cycle will have a much lower efficiency than the same LED driven at 250 mA and a 100% duty cycle, so your stated method will overestimate lumen production. That isn't to deny the role of heat, but heat isn't the whole story by any means.

For more details, find Newbie's posts and graphs (extensive) on the subject.

Yes, it is correct. Read the part where I mention going to the data sheet (which shows decreasing efficiency at higher power dissipation). If you read what I said (and I repeat here)

...you need to compute the average power at the particular duty cycle. From there you can go to the data sheet to determine what the efficiency is by looking at output vs. bias current...

The data sheet shows diminishing returns as you increase the current drive. Newbie has a great graph showing the difference between PWM and CC drive. Both show diminishing efficiency as you mononotically increase current to the LED.

 

[RustyKnee]

Yes, it is correct. Read the part where I mention going to the data sheet (which shows decreasing efficiency at higher power dissipation). If you read what I said (and I repeat here)



The data sheet shows diminishing returns as you increase the current drive. Newbie has a great graph showing the difference between PWM and CC drive. Both show diminishing efficiency as you mononotically increase current to the LED.

I can't see an efficiency versus drive current on newbies page that was linked above. I can see intensity against current. I can't see a plot newbie did that shows deminishing efficiency for the PWM method.

Not sure which datasheet you are looking as you don't say. I just checked the ones I have for the Seoul and Cree LEDs. Neither has a power vs intensity, they have current against intensity....which is different.

The intensity/current curves are a differnent shape for the 2 methods though. The intensity/current plot isn't itself an efficiency plot...but lookign at the results of the 2 curves you can work it out. efficiency in this case is lumens / watt.

Yes they both have the same efficiency at the start and end....but not in the middle. In the middle constant current is more efficient than PWM. The slope of the PWM plot is near enough linear, this indicates efficiency is about the same for all PWM average currents, not increasing or decreasing as average drive current is increased. eg in the middle of the PWM curve intensity is about half max for hald max current....half diveded by half is 1. the CC mthod has more than half intensity at half CC current....more than half divided by half is more than 1....more efficient.

Using the averaged pwm current on the data sheet intensity vs current plots doesn't give the correct answer.

Stu

 

[CM]

...The intensity/current curves are a differnent shape for the 2 methods though. The intensity/current plot isn't itself an efficiency plot...but lookign [sic] at the results of the 2 curves you can work it out. efficiency in this case is lumens / watt...

I put your statement in bold for emphasis. I think you made my point for me